GPT-5.6 Has Three Versions Now — Which One Are You Actually Using?
GPT-5.6 Has Three Versions Now — Which One Are You Actually Using?
OpenAI split its newest model into Sol, Terra, and Luna. Most explanations are written for developers. Here's what it means if you just open ChatGPT and type.
OpenAI began rolling out GPT-5.6 on July 9, 2026, and it isn't one model. It's three: Sol, Terra, and Luna. If you've seen those names and wondered which one you're supposed to pick, there's a short answer most articles skip.
For regular ChatGPT users, you're probably not picking at all. Here's what the three tiers actually are, which one you're using right now, and when the distinction matters to you.
Why OpenAI Split One Model Into Three
The reasoning is practical. Not every task needs the most powerful model available, and running a heavyweight model on a simple request is slow and expensive.
So instead of one giant model trying to be everything, OpenAI now offers three tiers along three axes: intelligence, speed, and cost. The number 5.6 identifies the generation; Sol, Terra, and Luna are durable capability tiers that each advance at their own pace.
The names come from the cosmos — sun, earth, moon — and they replace the older "mini" and "nano" labels that confused more people than they helped.
What Each Tier Is For
Sol — the peak
The flagship and most capable of the three. It's built for the hardest work: complex coding, long multi-step sessions, advanced agents, and research-grade analysis. It's the tier you want when quality matters more than cost.
Terra — the balanced
Strong performance at a lower price, positioned for large-scale business work like customer support, internal tools, and document analysis. Several analysts describe Terra as the sensible default for most professional workflows — close to Sol on many tasks without Sol's cost.
Luna — the fast one
The fastest and lowest-cost tier, aimed at light everyday tasks: summarizing, drafting, and routine automation. Despite the low price, it performs near the previous generation on several tests.
The Part Most Articles Leave Out
Here's what matters if you're a regular ChatGPT user rather than a developer.
GPT-5.6 Terra and GPT-5.6 Luna are not selectable in standard ChatGPT conversations. Sol powers the Medium, High, and Extra High reasoning options on eligible plans, and Sol Pro powers the Pro tier. Logged-out users don't get Sol at all.
So when you open ChatGPT and type a question on a paid plan, you're using Sol — not choosing between three models. The Sol/Terra/Luna decision is a developer and business decision, made through the API or Codex where cost per task actually shows up on a bill.
The Price Difference (And Why It Exists)
The gap is the whole reason the tiers exist. Per million tokens, input and output pricing runs roughly: Sol at $5 / $30, Terra at $2.50 / $15, and Luna at $1 / $6.
That's up to a fivefold difference on output between the top and bottom tier. For someone running thousands of requests through an application, picking the right tier for each task translates directly into a very different monthly bill. For someone asking a few questions a day in the app, it's invisible.
| Tier | Best for | Available in ChatGPT app? |
|---|---|---|
| Sol | Hardest reasoning, long sessions, complex coding | Yes — powers paid plan reasoning |
| Terra | Everyday business work at lower cost | No — not selectable in standard chat |
| Luna | High-volume simple tasks, speed-first work | No — not selectable in standard chat |
If You Do Build Things: Route, Don't Default
For anyone using the API or building workflows, the consensus in early analysis is that "always use Sol" is the wrong instinct.
Several reviewers recommend Terra as the practical default, because it lands within a few points of Sol on most benchmarks at roughly half the price. Sol earns its premium on genuinely hard agentic work — long multi-step tasks, complex terminal operations, anything where fewer retries and less supervision matter.
The more interesting pattern is mixing tiers inside one system. A support workflow might use Luna to detect intent and route messages, Terra to draft the responses, and Sol only for escalated cases that need deeper reasoning. That layered setup gets most of the quality at a fraction of the cost.
Quick guide to the three tiers:
- Using the ChatGPT app on a paid plan? You're on Sol
- Logged out? You don't have Sol access
- Building with the API? Consider Terra as your default
- Escalate to Sol only for genuinely hard tasks
- Use Luna for high-volume, simple, repetitive work
- Mixing tiers usually beats picking one
What This Signals About Where AI Is Going
Step back from the names for a second, because the structural change matters more than the release itself.
For years, the pattern was one flagship model that everyone used for everything — asking it to summarize a paragraph and to solve a research problem with the same machinery. That was simple to understand and wasteful in practice.
Splitting into tiers admits something obvious: most requests are easy, a few are hard, and pricing them identically doesn't make sense. Expect the same pattern from every major provider, because the economics push in one direction. The interesting consequence for regular users is that the cheap tiers keep getting good enough for ordinary work, which quietly makes AI cheaper over time rather than more expensive.
The Downsides Worth Knowing
The naming is the first problem. Sol, Terra, and Luna are more memorable than "mini" and "nano," but they don't tell you anything about capability order. Nothing in the words themselves says Sol is the strongest — you simply have to know. For a company trying to reduce confusion, that's a partial success at best.
Second, the split creates a decision where there used to be none. Developers now have to think about routing on every workflow, and getting it wrong in either direction costs something: over-provision and you burn budget, under-provision and you get worse results with more retries.
Third, and most relevant here — the announcement generated a lot of coverage that doesn't apply to most readers. If you're a regular ChatGPT user, "GPT-5.6 has three tiers" is technically true and practically irrelevant to your day. It's worth knowing so you can stop worrying you picked wrong, and then move on.
My Honest Take
Honestly, most people reading about Sol, Terra, and Luna don't need to do anything about it. If you type into ChatGPT on a paid plan, you're already on the strongest tier and there's no setting to adjust.
Where it genuinely matters is if you're building something — and there, the useful takeaway is the opposite of what the flagship marketing suggests. Terra being close to Sol at half the price makes the default choice cheaper, not more expensive. The tiering is arguably better news for budgets than for capability.
The broader shift is worth noticing though. AI companies are moving away from "one model to rule them all" toward routing work by difficulty. That's a more mature way to build, and it'll probably become normal across every major provider.
FAQ
Which GPT-5.6 model am I using in ChatGPT?
On eligible paid plans, you're using Sol — it powers the Medium, High, and Extra High reasoning options, with Sol Pro on the Pro tier. Terra and Luna aren't selectable in standard ChatGPT conversations.
Can I choose between Sol, Terra, and Luna?
Not in normal ChatGPT chats. The choice exists for developers and businesses accessing the models through the API or Codex, where cost per request matters.
Is Sol always the best option?
Not for everything. Sol leads on the hardest tasks, but Terra comes within a few points on many benchmarks at about half the price. For most workloads, analysts suggest defaulting to Terra and escalating to Sol only when needed.
What happened to the "mini" and "nano" model names?
OpenAI replaced them with the Sol, Terra, and Luna tier names. The company says the number identifies the generation while the names are durable capability tiers that advance at their own pace.
Do I get GPT-5.6 without logging in?
Logged-out users don't have access to Sol. In Business and Enterprise workspaces, admins may also control which models members can use.
The Bottom Line
GPT-5.6 arrived on July 9, 2026 as three tiers: Sol for hard work, Terra for balanced everyday use, Luna for speed and volume. If you use the ChatGPT app on a paid plan, you're on Sol and there's nothing to change.
If you build with the API, the practical advice runs the other way — start with Terra and escalate only when the task earns it.
Based on OpenAI's July 2026 rollout, OpenAI Help Center documentation, and contemporaneous analysis from independent reviewers. Pricing and availability may change — check OpenAI's official pages for current details. Researched with AI assistance and reviewed before publishing.
